Application of Model‐Informed Approaches in Neuroscience Drug Development and Regulatory Decision‐Making

Abstract

Model‐informed drug development (MIDD) approaches are being increasingly used in neuroscience drug development programs and regulatory decision‐making. In this work, the authors summarize a select set of cases in neurology, psychiatry, addiction, anesthesia, and analgesia, where MIDD approaches were utilized to address critical drug development questions or fill knowledge gaps. These examples include efficacy considerations such as biomarker‐clinical endpoint relationships, surrogate endpoints for accelerated approval, extrapolation of efficacy, and safety aspects such as informing driving studies. Additionally, the examples also inform optimal dosing regimen selection, dose adjustments in drug interactions, life cycle management decisions such as indication expansion, formulation switching, and informing patient population selection. Together, these examples offer insights into the utility of MIDD approaches to streamline drug development and opportunities to save costs and resources.

Introduction

The complexity of the central nervous system (CNS), coupled with the often slow and heterogeneous progression of neurological and psychiatric disorders, has historically posed significant hurdles in developing novel therapies for these conditions. This is empirically reflected in clinical trial outcomes, where CNS drug development programs have one of the lowest success rates of any therapeutic area. A landmark academic analysis found the probability of success from Phase 1 to approval for neurology assets was only 6.3%. ¹ This trend is corroborated by more recent, ongoing industry analyses; for instance, a comprehensive report covering the 2011–2020 period confirmed this challenge, finding the success rate for CNS therapies remained low at 5.9%. ² Although many factors can lead to program failure, and a detailed discussion on the cause of trial failure is beyond the scope of this paper, this study found clinical programs incorporating biomarkers for patient stratification were associated with higher success rates. These challenges highlight the need for quantitative approaches that integrate diverse sources of information to optimize dose selection, support patient selection, improve trial design, and strengthen regulatory decision‐making.

Model‐informed drug development (MIDD) approaches are defined as the use of computational modeling and simulation (M&S) methods that can include and integrate nonclinical data, clinical data, prior information, and knowledge (e.g., drug and disease characteristics) to generate evidence. ² These approaches involve stakeholders across academia, industry, and regulatory agencies in multiple disciplines including clinical, biostatistics, and clinical pharmacology. The impact of the application of such quantitative approaches to inform approval and labeling decisions has been reported previously, ³ , ⁴ , ⁵ , ⁶ and industry groups also published extensively on this topic. ⁷ , ⁸ , ⁹ , ¹⁰ Building on its long history of demonstrated value, the US Food and Drug Administration (FDA) continues to champion MIDD approaches and their potential to modernize the drug development and regulatory evaluation process. ¹¹ This endorsement has been further solidified through concrete actions and initiatives aimed at integrating these approaches into the regulatory lifecycle. ¹² A key example is the MIDD Paired Meeting Program, which created a formal pathway for collaboration between drug sponsors and the agency. ¹³

The value proposition of MIDD extends beyond improved scientific understanding and regulatory alignment; it offers a business case for a more time‐ and cost‐efficient development paradigm. For instance, Sahasrabudhe et al ¹⁴ reported an annualized average savings of approximately 10 months of cycle time and $5 million per program due to sample size reduction, waivers of clinical trials, and “No‐Go” decisions for conducting a trial. In a similar vein, a survey of industry stakeholders also highlighted significant benefits such as up to 2 years in time and up to $70 million in cost‐savings from participation in the MIDD paired‐meeting pilot program ¹⁵ due to many reasons including smaller (reduced) trials, simulated outcomes replacing the need for clinical trials, avoided using resources to test the wrong dose levels, and evaluating PKPD on less costly but validated biomarkers to demonstrate proof of efficacy and choosing the optimal doses to test in subsequent trials.

This work is aimed at summarizing the impact of MIDD approaches in informing drug development and regulatory decisions for products submitted to the Office of Neuroscience at the Center for Drug Evaluation and Research (CDER), FDA. The Office of Neuroscience evaluates products in neurology, psychiatry, anesthesiology, addiction medicine, and pain medicine areas.

Methods

The authors screened regulatory reviews of new drug applications (NDAs) and biologic license applications (BLAs) submitted to the Office of Neuroscience over the last 10 years. Candidate submissions included NDAs, supplemental applications, and post‐marketing applications. The authors identified instances where model‐informed approaches had a significant impact on the development, approval, or directions of use of a drug or biologic. The selected cases were intended as illustrative examples of the diverse applications of MIDD in neuroscience drug development and regulatory decision‐making and should not be interpreted as a systematic review or an exhaustive assessment of all neuroscience submissions. Cases were considered to have a significant impact if MIDD contributed to one or more of the following criteria: (1) reduction in sample size, (2) replacement of a clinical study through simulation, (3) dose selection or dose optimization, (4) support pediatric extrapolation, (5) support for biomarker qualification or surrogate endpoints (reasonably likely to predict clinical benefit), or (6) informing regulatory approval, labeling, or other key regulatory decisions.

Various model‐informed approaches were categorized under broad categories as they serve a specific purpose:

Population pharmacokinetic (PPK) models assess the impact of various intrinsic factors (e.g., hepatic impairment and renal impairment) or extrinsic factors (e.g., food effect and drug–drug interactions) on the variability in drug pharmacokinetics. Drug pharmacokinetics are typically characterized using C_max (maximum plasma/serum concentrations), C_trough (concentrations immediately before the next dose is administered), C_avg (average concentrations during a dosing interval at steady state), and AUC (area under the curve).

Physiologically based pharmacokinetic (PBPK) models include terms to describe anatomical and physiological processes involved in drug disposition. In neuroscience drug development, PBPK models may be helpful for describing drug penetration across the blood–brain barrier and predicting central nervous system drug exposure in addition to concentrations in plasma.

Exposure–response (E‐R) models describe the relationship between a drug exposure metric (e.g., C_max, C_trough, C_avg, AUC, time above a threshold concentration, etc.) and clinical outcome. This tool can be invaluable in neuroscience applications, where the therapeutic response to a drug may be delayed due to complex biological processes involved. Such an approach can incorporate relevant CNS‐specific biomarkers (such as cerebrospinal fluid markers, electrophysiological recordings, etc.).

Quantitative systems pharmacology (QSP) models integrate multiple biological processes and disease mechanisms. Such models may integrate multiple scales of data, from molecular to clinical outcomes. This approach may allow characterization of complex interactions between drugs, biological targets, and physiological systems.

Results

Representative cases where model‐informed approaches had a significant impact in approval are discussed below. (Table 1).

Table 1.

Examples of the Application of MIDD Principles in Drug Development and Regulatory Decision‐Making

MIDD Topic	Therapeutic Indication	Drug or Biologic	Methods	Impact
PK bridging to expand indication for a new formulation	Schizophrenia and bipolar I disorder	Risperidone long‐acting injection	PPK	Supported the approval of the new formulation without an additional efficacy clinical study
Support switching between formulations, dosing windows, and missed dosing for long‐acting injectables	Schizophrenia and schizoaffective disorder	Paliperidone palmitate	PPK	Supported formulation switching, flexible dosing recommendations, and missed‐dose labeling without additional clinical studies
Pediatric extrapolation	Epilepsy (partial onset seizures)	Brivaracetam, eslicarbazepine acetate, lacosamide, pregabalin, and perampanel	PPK	Supported pediatric efficacy extrapolation without dedicated pediatric efficacy studies Supported pediatric efficacy extrapolation
	Schizophrenia	Lurasidone and brexpiprazole	PPK
	Bipolar I disorder	Asenapine	PPK
Surrogate endpoints (biomarkers as reasonably likely to predict clinical benefit)	SOD1‐amyotrophic lateral sclerosis	Tofersen	PPK, PKPD, E‐R	Supported the use of relevant biomarkers as surrogate endpoints (i.e., reasonably likely to predict clinical benefit) for accelerated approval
	Alzheimer's disease	Aducanumab, lecanemab	PPK, PKPD, E‐R
Supporting dosing regimen not directly evaluated in clinical trials	Duchenne muscular dystrophy	Givinostat	PPK, PKPD, E‐R	Supported the approval of a dosing regimen not directly evaluated in clinical trials without an additional efficacy study
	Tardive dyskinesia	Valbenazine	PPK, E‐R
	Generalized myasthenia gravis	Rozanolixizumab	PPK
	Opioid overdose reversal	Naloxone	PPK
Support approval of emergency opioid overdose treatment	Emergency treatment of opioid overdose	Nalmefene	PPK, PKPD	Supported approval by informing dose selection and pediatric extrapolation (age 12 to 17 years).
PBPK models to support dosing without conducting clinical studies	Spinal muscular atrophy	Risdiplam	PBPK	Supported dosing recommendations without conducting additional clinical PK studies
	Schizophrenia and bipolar I disorder	Cariprazine	PBPK
Driving studies	Postpartum depression	Zuranolone	E‐R models	Supported labeling recommendations regarding next‐day driving restrictions

PPK, population pharmacokinetic; PKPD, pharmacokinetic pharmacodynamic; E‐R, exposure–response; PBPK, physiologically based pharmacokinetic; DP, disease progression; QSP, quantitative systems pharmacology; SOD1, superoxide dismutase 1 gene.

PK Bridging to Expand the Indication for a New Formulation

Uzedy (risperidone long‐acting injection, LAI), administered once monthly, Q1M (50, 75, 100, and 125 mg) or once every 2 months, Q2M (100, 150, 200, and 250 mg), was initially approved for the treatment of schizophrenia in adults. The Applicant sought to expand the use of Uzedy for the treatment of bipolar I disorder based on an MIDD approach. A model‐based PK bridge was employed with Risperdal Consta (25, 37.5, and 50 mg once every 2 weeks, Q2W) as the listed drug (approved for treatment of schizophrenia and bipolar I disorder). Previously developed and validated population PK models for both Risperdal Consta and Uzedy were used to simulate steady‐state exposure under their respective dosing regimens. The published population PK model for Risperdal Consta was verified using clinical data generated by the Applicant for Uzedy to support the reliability and fitness of the model for PK comparison between Risperdal Consta and Uzedy. The Uzedy model was established and verified during its original NDA submission. Comparative analyses demonstrated that the key PK metrics (C_max, C_trough, and AUC at steady state) for Uzedy Q1M (50, 75, or 100 mg, but not 125 mg), but not Uzedy Q2M, were comparable to those of Risperdal Consta Q2W (25, 37.5, or 50 mg), with the geometric mean ratios within the 0.8 to 1.25 range. These results supported the adequacy of the PK bridge between the two products. These analyses supported the approval of Uzedy Q1M (50, 75, or 100 mg) for bipolar I disorder without requiring a clinical study. ¹⁶

Support for Switchability Between Formulations, Dosing Windows, and Missed Dose Management for a Long‐Acting Injectable

In clinical practice, patients maintained on a stable dose of one formulation may require transition to an alternative formulation, necessitating careful consideration of therapeutic equivalence and safety. Erzofri, is a once‐monthly, extended‐release long‐acting injectable (LAI) formulation of paliperidone palmitate, relied on the Agency's previous findings of safety and effectiveness of an approved once‐monthly LAI product, Invega Sustenna (listed drug, LD) for the treatment of schizophrenia and schizoaffective disorder in adults. The new formulation, Erzofri, offers an initiation regimen that requires only a single loading dose on Day 1, compared to two loading doses for the LD. ¹⁷

A PPK model was developed using pooled data from clinical studies including patients treated with both Erzofri and LD. ¹⁸ , ¹⁹ The model was used to simulate various clinical scenarios in which patients at steady state on Invega Sustenna were switched to or from the same nominal dose of Erzofri at the next scheduled injection. Simulations demonstrated that paliperidone plasma concentration–time profiles were comparable following the switch. These findings support that clinical efficacy and safety would be preserved when transitioning patients from Invega Sustenna to Erzofri and vice versa. Modeling analyses also supported flexibility of administration of Erzofri 7 days before or after the scheduled monthly administration if a dose is missed. Model‐based simulations were performed to support dosage recommendations for patients who missed a dose of Erzofri for different time periods (e.g., 4 to 6 weeks, more than 6 weeks to less than 6 months, and longer than 6 months after last injection). These efforts highlight the significance of model‐informed approaches to support switchability between the formulations, dosing windows, and missed dose management for LAI products, which would otherwise require clinical studies that are impractical to conduct.

Deriving Policy to Support Pediatric Extrapolation

Model‐informed approaches played an important role in supporting pediatric efficacy extrapolation to enable faster access to treatments for partial‐onset seizures (POS) and schizophrenia in the pediatric population. The scientific basis of extrapolation between adults and pediatric patients for epilepsy relied on the demonstration of similar disease pathophysiology, ²⁰ similarity in response to treatment (reduction in seizure frequency), and similar E‐R relationship. ²¹ Analysis of trials evaluating drugs for schizophrenia and bipolar I disorder that share a similar mechanism of action (i.e., D2 receptor antagonism or partial agonism, 5‐HT1A receptor partial agonism, and/or 5‐HT2A receptor antagonism) suggested similar symptomatic changes following acute relapse and similar E‐R relationship in adults and pediatric patients with manic/mixed episodes associated with bipolar I disorder and schizophrenia. ²² These findings confirmed that extrapolation of efficacy from adult to pediatric patients can be justified.

The pediatric extrapolation framework was based on a collective effort between industry, academia, and the FDA in which a systematic quantitative analysis of data with several drugs with a variety of putative mechanisms of action was conducted. Under this framework, model‐informed analyses were used to bridge adult efficacy data to pediatric populations, thereby reducing or eliminating the need for dedicated pediatric efficacy trials in some instances.

These analyses supported pediatric efficacy extrapolation and contributed to the approval of six therapies for the treatment of epilepsy in the pediatric population without the need for dedicated efficacy trials.

Establishing a Biomarker as a Reasonably Likely Surrogate Endpoint

Qalsody (tofersen) was approved under the accelerated approval pathway based on the determination that SOD1‐ALS is a serious condition with no available therapies, and its effect on plasma neurofilament light chain (NfL), which was considered a surrogate endpoint reasonably likely to predict clinical benefit. Model‐informed analyses were used to evaluate the relationship between plasma NfL and clinical outcomes. A meta‐analysis of published data demonstrated a consistent association between plasma NfL levels and disease progression. The relationship between tofersen‐driven NfL reduction and changes in clinical endpoints was evaluated using correlation and causal inference approaches. These analyses demonstrated that reduction in plasma NfL is associated with slower rates of clinical decline, providing quantitative support for the use of plasma NFL as a surrogate endpoint that is reasonably likely to predict clinical benefit in SOD1‐ALS patients. ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ These analyses provided quantitative support for plasma NfL as a reasonably likely surrogate endpoint and contributed to the accelerated approval decision for Qalsody.

A cross‐program meta‐analysis of anti‐amyloid therapies demonstrated a consistent relationship between the magnitude of amyloid plaque reduction and clinical outcomes at the group level, supporting the use of beta‐amyloid plaque reduction as a reasonably likely surrogate endpoint for accelerated approval. These analyses provided support for including changes in amyloid plaque as a marker for various regulatory decisions. For Leqembi (lecanemab), an integrated PK‐amyloid plaque reduction model was used to evaluate a reduced‐frequency maintenance dosing regimen (once every 4 weeks, Q4W), with simulations predicting minimal differences in amyloid burden and clinical outcomes compared with continued biweekly dosing. ²⁸ For Kisunla (donanemab), a PK‐amyloid model was used to characterize amyloid re‐accumulation following treatment discontinuation, informing evaluation of a biomarker‐guided dosing approach. ²⁹ These analyses informed regulatory decision‐making regarding dosing optimization and the use of amyloid plaque reduction as a biomarker supporting accelerated approval and post‐approval dosing optimization strategies for Leqembi and Kisunla.

Support the Approval of Dosing Regimen not Directly studied in Clinical Trials

Rystiggo (rozanolixizumab) is a neonatal Fc receptor (FcRn) blocker approved for the treatment of generalized myasthenia gravis (gMG) in adult patients. The pivotal Phase 3 study evaluated the efficacy and safety of rozanolixizumab using a weight‐tiered dosing regimen where patients received doses approximating 7 or 10 mg/kg. To simplify administration and to reduce the potential for dosing errors in clinical practice, a fixed dose regimen was proposed. Population PK/PD modeling and simulation were used to evaluate a modified dosing regimen (420 mg for patients <50 kg, 560 mg for patients ≥50 kg to <100 kg, and 840 mg for patients ≥100 kg). The simulations demonstrated that the proposed fixed‐dose regimens provided exposures comparable to those achieved with the studied weight‐based regimen. ³⁰ These analyses supported approval of a simplified fixed‐dose regimen by demonstrating exposure comparable to the clinically studied weight‐based regimen, eliminating the need for additional clinical study. ³¹

A similar approach was used for Duvyzat (givinostat), approved for the treatment of Duchenne Muscular Dystrophy (DMD). The pivotal Phase 3 study included a body weight‐based dosing regimen comprising nine body weight categories. A simplified dosing regimen with four body weight categories for body weight‐based dosing was evaluated to simplify dosage administration. PK, PK‐efficacy (Total North Star Ambulatory Assessment Score, NSAA), and PK‐Safety (platelet count) simulations demonstrated comparable outcomes between the studied and the unstudied simplified regimens. ³² These analyses supported approval of a simplified body weight‐based dosing regimen without requiring additional clinical evaluation of the modified regimen. ³³

Model‐informed approaches have also been applied in settings where clinical studies are not feasible. An example is the development of a 10‐mg naloxone autoinjector as a medical countermeasure against weaponized high‐potency synthetic opioids like fentanyl and carfentanil. FDA's independent mechanistic PK‐PD modeling validated sponsor claims that 10 mg naloxone achieved higher rescue rates compared to lower doses (2 mg) for fentanyl and carfentanil overdoses. ³⁴ , ³⁵ Modeling demonstrated that prophylactic administration 5–30 min prior to synthetic opioid exposure would prevent >90% cardiac arrests. Further, simulations emphasized the importance of early dosing (when respiratory ventilation drops to 90% of baseline), which significantly improved outcomes, particularly for high‐dose carfentanil scenarios. These analyses supported the approval of 10‐mg naloxone autoinjector for emergency treatment and prophylaxis in patients ≥12 years in scenarios involving high‐potency synthetic opioids without requiring efficacy trials that would be infeasible to conduct.

Another example is Ingrezza (valbenazine), initially approved in 2017 at doses of 40 and 80 mg for the treatment of tardive dyskinesia in adults. Approval of a 60‐mg dose strength was subsequently sought based on real‐world data and E‐R findings without conducting a clinical study at that dose level. An E‐R model for Abnormal Involuntary Movement Scale (AIMS) was developed to characterize the efficacy for the 60‐mg dose. This dose was expected to be well‐tolerated based on the overall safety profile across the dose range. The mean change from baseline in AIMS total score at Week 6 for the 60 mg dose is included in the approved label. ³⁶ These analyses supported the approval of the 60‐mg dose by demonstrating the expected efficacy and safety profile without conducting an additional dose‐finding clinical study.

Use of Systems Pharmacology Modeling to Support Approval of Emergency Opioid Overdose Treatment

Zurnai (nalmefene 1.5 mg autoinjector) was developed for emergency opioid overdose treatment in community settings. FDA raised concerns about intramuscular/subcutaneous nalmefene formulations, including slow onset of action (5–15 min per nalmefene injection label), appropriateness for community use without titration, and potential for protracted withdrawal symptoms. The autoinjector formulation aimed to address these limitations while providing an alternative to intranasal products. FDA's systems pharmacology model translated healthy volunteer PK‐PD data (fentanyl‐induced respiratory depression model) to community scenarios. Simulations in virtual chronic opioid users assessed cardiac arrest prevention and brain hypoxia time for fentanyl/carfentanil overdoses. PPK modeling supported pediatric extrapolation (12–17 years). Repeat‐dose simulations using nonparametric superposition confirmed safety coverage versus approved IV nalmefene regimens. The product was approved for patients ≥12 years. ³⁷ Model‐informed evidence was pivotal, demonstrating onset comparable to naloxone 4 mg intranasal (reduced cardiac arrest from 52% to <10% for fentanyl), no titration requirement, and acceptable repeat‐dose safety. ³⁸

These analyses supported the approval of Zurnai by demonstrating clinically meaningful reversal of opioid induced respiratory depression and informing dosing, safety, and pediatric extrapolation strategies.

Use of PBPK Models to Support Dosing and Minimizing the Need for Drug Interaction Studies

PBPK modeling can support dose selection and evaluate drug–drug interaction risk in settings where clinical studies are limited or infeasible.

Evrysdi (risdiplam) is an orally administered SMN2 splicing modifier approved for the treatment of spinal muscular atrophy (SMA) across a broad pediatric age range, including neonates. A key regulatory challenge was establishing appropriate dosing for infants younger than 2 months, in whom PK or safety data were not available. PBPK modeling was used to support dosing recommendations for infants ≤ 2 months of age. The model incorporated age‐dependent physiological changes, including body size and maturation of drug‐metabolizing enzymes, primarily flavin‐containing monooxygenases (FMO1 and FMO3) and CYP3A. To address uncertainty in FMO3 ontogeny, multiple PBPK scenarios representing a range of plausible maturation profiles were evaluated. Simulations predicted a range of exposures in neonates initiating treatment at birth, with predicted exposures similar to those observed and tolerated in slightly older patients. ³⁹ These analyses defined the bounds of expected exposure and supported benefit–risk assessment, enabling approval of risdiplam for use from birth onward without requiring direct clinical data in the youngest patients.

A complementary application of PBPK modeling is illustrated by Vraylar (cariprazine), an atypical antipsychotic metabolized primarily by CYP3A4. Clinical DDI studies were limited by the long half‐lives of cariprazine and its active metabolites, and insufficient treatment duration may underestimate interaction effects at steady state. PBPK modeling was therefore used to predict the impact of strong and moderate CYP3A4 inhibitors on total cariprazine exposure, including contributions from its active metabolites. Simulations predicted increases in AUCτ of approximately 5.5‐ and 3‐fold with strong and moderate inhibitors, respectively, and were used to evaluate alternative dose adjustment strategies. To improve patient adherence, lower dose strengths (0.5 and 0.75 mg) were developed to support daily administration instead of less frequent administration (e.g., every 2 days or every 3 days). PBPK simulations demonstrated that daily administration of these lower doses would provide exposures comparable to those achieved with the approved intermittent dosing schedules in the presence of CYP3A4 inhibitors. ⁴⁰ These analyses informed labeling recommendations and supported approval of revised dosing recommendations without requiring additional clinical DDI studies. ⁴¹

Use of ER Models to Inform Next‐Day Ability to Safely Operate Motor Vehicles

Zurzuvae (zuranolone) is approved for the treatment of postpartum depression (PPD) in adults. In two motor‐vehicle driving simulation studies, the driving ability of healthy adults was impaired in a dose‐dependent manner following repeat nighttime administration of zuranolone. A need to provide instructions on when patients could safely operate a motor vehicle the next day morning was identified. The driving simulation studies evaluated the driving ability impairment at a pre‐specified timepoint after dosing. To inform this driving ability impairment at other time points, ER analysis using information at low dose and therapeutic dose was conducted. Effects at various time points after dosing were simulated. ⁴² The label indicates that Zurzuvae may impair driving due to CNS depressant effects and recommends avoiding driving or hazardous activities for at least 12 h after administration. ⁴³ These analyses informed labeling recommendations regarding next‐day driving by characterizing the relationship between Zurzuvae exposure and driving impairment without conducting additional clinical studies.

Disease Progression Modeling to Inform Patient Selection Criteria for Rare Disease Clinical Trials

Disease progression modeling is a tool that can help characterize natural history and inform patient selection criteria for rare disease clinical trials. Disease progression analysis can help when comparing on‐treatment findings with the natural history of the disease. For example, quantifying associations between genetic mutations and loss of ambulation (LoA) among males diagnosed with childhood‐onset dystrophinopathy is considered important for understanding variation in disease progression. ⁴⁴ , ⁴⁵

Findings from these analyses suggested that patients with certain mutations have slower disease progression than others and should be considered in clinical trial design. Such modeling efforts can also help when conducting propensity score matching analysis comparing long‐term treatment data versus natural history data. ³² , ³³ These analyses illustrate how disease progression modeling can support rare disease drug development by informing patient selection, improving clinical trial design, and evaluating treatment effects and comparing to natural history data.

Discussion

The regulatory landscape for model‐informed approaches has undergone a remarkable transformation, evolving from niche applications to a comprehensive, institutionalized framework that spans the entire drug development lifecycle. Model‐informed approaches began emerging in the field in the late 1990s with prominent milestones including the first population pharmacokinetic guidance document in 1999 and the first ER analysis guidance document in 2003. The End of Phase 2A meeting program was the first attempt by the Agency to provide a structured meeting approach to address model‐informed approaches.

The recognition of model‐informed approaches as a key regulatory decision tool to advance drug development under the sixth iteration of the Prescription and Drug User Fee Act (PDUFA VI) has been a key driver for holistic and integrative advancement of MIDD. The MIDD Paired Meeting Program, initiated as a pilot in 2018 and later as a formal meeting program under PDUFA VII, has demonstrated substantial real‐world impact beyond the theoretical promise of these approaches. The demand for MIDD paired meetings has grown significantly since the program's inception, reflecting increasing industry recognition of the value of these approaches. Notably, the program has resulted in multiple regulatory approvals that would have been challenging or impossible without model‐informed evidence. ⁴⁶ These successes validate the business case for MIDD, with survey results from the International Consortium for Innovation and Quality in Pharmaceutical Development confirming industry‐reported savings of $30 to $70 million and up to 2 years per program.

MIDD is a powerful quantitative framework that integrates evidence from preclinical studies and clinical trials across all stages of development to inform drug development and regulatory decision‐making. The success of MIDD application to address a drug development or regulatory question depends on high‐quality data relevant to the key question, appropriate model validation, and a clearly defined context of use. Because MIDD relies on the available evidence, the quality, completeness, and representativeness of the underlying data directly influence the reliability of model development and the resulting inferences. Consequently, limitations in the available data may increase uncertainty in model predictions and their interpretation. In addition, applications in disease areas where the underlying pathologic mechanisms (e.g., many psychiatry indications) are poorly understood may be associated with greater uncertainty in the model predictions compared to applications where the indication is well‐understood. Finally, successful implementation of MIDD requires additional training and expertise due to complex technical methods, software knowledge, and programming expertise.

This dedicated form of regulatory engagement on MIDD issues in specific drug development programs has been complemented by the establishment of parallel infrastructure to support broader adoption. The Fit‐For‐Purpose (FFP) initiative addresses a critical need for reusable, publicly available quantitative tools that can be applied across multiple development programs. The public availability of these determinations facilitates greater utilization and promotes efficiency through shared resources. ⁴⁷ The 2024 establishment of the Quantitative Medicine Center of Excellence (QM CoE) as a CDER‐wide enterprise represents a structural evolution in how FDA organizes expertise to support MIDD. ⁴⁸ This interdisciplinary network spans multiple participating offices across CDER, providing coordinated expertise that transcends traditional organizational boundaries. The QM CoE serves not only to facilitate individual product reviews but also to drive continuous evolution of methodologies and ensure consistent application of quantitative medicine principles across therapeutic areas. Lastly, the recent harmonization of the ICH M15 guideline on General Principles for MIDD provides an international framework that extends beyond FDA to global regulatory authorities. ⁴⁹ As quantitative methods continue to mature and regulatory experience grows, MIDD is expected to play an increasingly important role in integrating diverse sources of evidence to support neuroscience drug development.

Conclusions

The case examples presented in this manuscript illustrate how model‐informed approaches played an important role in improving efficiency in neuroscience drug development and regulatory decision‐making. These applications span a wide range of regulatory contexts, including pediatric extrapolation (e.g., POS and schizophrenia), biomarker‐endpoint relationships (e.g., plasma NfL in SOD1 ALS and beta amyloid plaque reduction in Alzheimer's disease), approval of dosing regimens not directly studied in clinical trials (e.g., rozanolixizumab, givinostat, valbenazine, and lecanemab maintenance dosing), formulation bridging (e.g., risperidone LAI and paliperidone LAI), safety assessments that would be impractical through traditional empirical approaches (e.g., naloxone and nalmefene for opioid overdose and zuranolone driving impairment), and characterization of complex ER relationships in heterogeneous diseases (e.g., anti‐amyloid therapies for Alzheimer's disease, Duchenne muscular dystrophy, and amyotrophic lateral sclerosis).

As the regulatory framework continues to evolve, with ongoing development of guidance, qualification of reusable tools through the FFP initiative, and coordination through the QM CoE, the potential for MIDD to transform neuroscience drug development will only increase. The continued collaboration between industry, academia, and regulatory authorities in advancing quantitative approaches will be critical to addressing the unique challenges of CNS drug development and ultimately improving outcomes for patients.

Disclosures

This reflects the views of the authors and should not be construed to represent FDA's views or policies.

Conflicts of Interest

The authors declare that they have no conflicts of interest to disclose.

Data Availability Statement

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.